{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:YXJP6AO6DG4PNXDYAW3I3QHN7K","short_pith_number":"pith:YXJP6AO6","canonical_record":{"source":{"id":"2506.21560","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-11T22:49:42Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dda9590af40732432f7fb1588e84475a11d7c6de9eb75b130030ee257c6ba27a","abstract_canon_sha256":"6c4c86a06eae47a4e7b50c46ea05f648ef60b69841a745dc8a23408721503557"},"schema_version":"1.0"},"canonical_sha256":"c5d2ff01de19b8f6dc7805b68dc0edfabb6da353d3f2dabd39df8c537cf23fc1","source":{"kind":"arxiv","id":"2506.21560","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.21560","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"arxiv_version","alias_value":"2506.21560v2","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.21560","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"pith_short_12","alias_value":"YXJP6AO6DG4P","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"pith_short_16","alias_value":"YXJP6AO6DG4PNXDY","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"pith_short_8","alias_value":"YXJP6AO6","created_at":"2026-07-05T11:44:07Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:YXJP6AO6DG4PNXDYAW3I3QHN7K","target":"record","payload":{"canonical_record":{"source":{"id":"2506.21560","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-11T22:49:42Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dda9590af40732432f7fb1588e84475a11d7c6de9eb75b130030ee257c6ba27a","abstract_canon_sha256":"6c4c86a06eae47a4e7b50c46ea05f648ef60b69841a745dc8a23408721503557"},"schema_version":"1.0"},"canonical_sha256":"c5d2ff01de19b8f6dc7805b68dc0edfabb6da353d3f2dabd39df8c537cf23fc1","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:44:07.936286Z","signature_b64":"UmYq6a+jHVJxnJNptY5DHrNbSPPG1sUP55lEcfdQDn4K3Yc2hpGd+HUH8LccruS3ls/JUdFDGQ+g3ekXNHvgBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c5d2ff01de19b8f6dc7805b68dc0edfabb6da353d3f2dabd39df8c537cf23fc1","last_reissued_at":"2026-07-05T11:44:07.935802Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:44:07.935802Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2506.21560","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:44:07Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LoLZNU1vPbnFy+Z3qhOyJiAAqcTU2P6QW7u6SmRDrftgTictKF8zbGaWrecIcKPlzcUU4mFki5dll2qQrUWGAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-23T05:08:05.613315Z"},"content_sha256":"60e74b86a841b24fc930cd96a26d8542d54319ed5d0d1ed25f6d4d723b165401","schema_version":"1.0","event_id":"sha256:60e74b86a841b24fc930cd96a26d8542d54319ed5d0d1ed25f6d4d723b165401"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:YXJP6AO6DG4PNXDYAW3I3QHN7K","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Reinforcement learning fine-tuning of language model for instruction following and math reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Geo Zhang, Yifu Han","submitted_at":"2025-06-11T22:49:42Z","abstract_excerpt":"This study investigates the effectiveness of reinforcement learning (RL) fine-tuning techniques on a compact language model (Qwen2.5-0.5B Base) for two challenging tasks: instruction following and mathematical reasoning. We compare supervised fine-tuning (SFT), Direct Preference Optimization (DPO) using preference-labeled data, and Reinforce Leave-One-Out (RLOO) with reward models. Our experiments show that RLOO with DeBERTa reward modeling achieves the best alignment, while DPO provides strong and consistent results. For math reasoing tasks, synthetic data augmentation and best-of-N sampling "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.21560","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2506.21560/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:44:07Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"NbGfwFDzGF/34uMKxvTgsE3x+pO7Xs1vT7RmvrPtJG8pJhcytWkDkaDgddruDonghO7UFq0/C6UHRxqJ3jbJBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-23T05:08:05.613644Z"},"content_sha256":"9eabba749d24acc5c7052b2e2dc9b874692dc7e92577c109afa6a38d8cf1c05e","schema_version":"1.0","event_id":"sha256:9eabba749d24acc5c7052b2e2dc9b874692dc7e92577c109afa6a38d8cf1c05e"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/YXJP6AO6DG4PNXDYAW3I3QHN7K/bundle.json","state_url":"https://pith.science/pith/YXJP6AO6DG4PNXDYAW3I3QHN7K/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/YXJP6AO6DG4PNXDYAW3I3QHN7K/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-23T05:08:05Z","links":{"resolver":"https://pith.science/pith/YXJP6AO6DG4PNXDYAW3I3QHN7K","bundle":"https://pith.science/pith/YXJP6AO6DG4PNXDYAW3I3QHN7K/bundle.json","state":"https://pith.science/pith/YXJP6AO6DG4PNXDYAW3I3QHN7K/state.json","well_known_bundle":"https://pith.science/.well-known/pith/YXJP6AO6DG4PNXDYAW3I3QHN7K/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:YXJP6AO6DG4PNXDYAW3I3QHN7K","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"6c4c86a06eae47a4e7b50c46ea05f648ef60b69841a745dc8a23408721503557","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-11T22:49:42Z","title_canon_sha256":"dda9590af40732432f7fb1588e84475a11d7c6de9eb75b130030ee257c6ba27a"},"schema_version":"1.0","source":{"id":"2506.21560","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.21560","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"arxiv_version","alias_value":"2506.21560v2","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.21560","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"pith_short_12","alias_value":"YXJP6AO6DG4P","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"pith_short_16","alias_value":"YXJP6AO6DG4PNXDY","created_at":"2026-07-05T11:44:07Z"},{"alias_kind":"pith_short_8","alias_value":"YXJP6AO6","created_at":"2026-07-05T11:44:07Z"}],"graph_snapshots":[{"event_id":"sha256:9eabba749d24acc5c7052b2e2dc9b874692dc7e92577c109afa6a38d8cf1c05e","target":"graph","created_at":"2026-07-05T11:44:07Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2506.21560/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"This study investigates the effectiveness of reinforcement learning (RL) fine-tuning techniques on a compact language model (Qwen2.5-0.5B Base) for two challenging tasks: instruction following and mathematical reasoning. We compare supervised fine-tuning (SFT), Direct Preference Optimization (DPO) using preference-labeled data, and Reinforce Leave-One-Out (RLOO) with reward models. Our experiments show that RLOO with DeBERTa reward modeling achieves the best alignment, while DPO provides strong and consistent results. For math reasoing tasks, synthetic data augmentation and best-of-N sampling ","authors_text":"Geo Zhang, Yifu Han","cross_cats":["cs.AI"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-11T22:49:42Z","title":"Reinforcement learning fine-tuning of language model for instruction following and math reasoning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.21560","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:60e74b86a841b24fc930cd96a26d8542d54319ed5d0d1ed25f6d4d723b165401","target":"record","created_at":"2026-07-05T11:44:07Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"6c4c86a06eae47a4e7b50c46ea05f648ef60b69841a745dc8a23408721503557","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-11T22:49:42Z","title_canon_sha256":"dda9590af40732432f7fb1588e84475a11d7c6de9eb75b130030ee257c6ba27a"},"schema_version":"1.0","source":{"id":"2506.21560","kind":"arxiv","version":2}},"canonical_sha256":"c5d2ff01de19b8f6dc7805b68dc0edfabb6da353d3f2dabd39df8c537cf23fc1","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c5d2ff01de19b8f6dc7805b68dc0edfabb6da353d3f2dabd39df8c537cf23fc1","first_computed_at":"2026-07-05T11:44:07.935802Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:44:07.935802Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"UmYq6a+jHVJxnJNptY5DHrNbSPPG1sUP55lEcfdQDn4K3Yc2hpGd+HUH8LccruS3ls/JUdFDGQ+g3ekXNHvgBA==","signature_status":"signed_v1","signed_at":"2026-07-05T11:44:07.936286Z","signed_message":"canonical_sha256_bytes"},"source_id":"2506.21560","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:60e74b86a841b24fc930cd96a26d8542d54319ed5d0d1ed25f6d4d723b165401","sha256:9eabba749d24acc5c7052b2e2dc9b874692dc7e92577c109afa6a38d8cf1c05e"],"state_sha256":"3612bb809a3467f8f4b4d1380f0d81cc0d0cd3e01041737f29af1a2100d5da4e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3Bz6itqDtjLcCX2U9LwskcCV8ugNHh3QhV/9wlQ4GodABTEOsSKWf/if0iqHMSrpmDJTw3+OfmHdLtvgWiwODA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-23T05:08:05.616963Z","bundle_sha256":"bdaeb6389015ea5bfd87a19b42f8e0511f9951a8691c42e9dda24bb3ba18515f"}}